EVIDENCE
UPDATED

What is done, on which board, and how we know.

This is the AGS-M1's record: the field test, the gates, and a dated log in which each line names its board and the class of evidence it earned. The AGS-G1 is in design, and its progress is on its own page.

FIELD-TESTED

Done outside the lab, in a real magnetic environment.

MEASURED

Done on the bench, with the raw data kept beside the report.

IN DESIGN

Designed or analysed, with no hardware result yet. Its numbers are design values, targets or estimates.

OPEN

Committed but not done. Listed so nobody mistakes silence for a result.

Field record

FIELD-TESTED
Board
Rev B, S/N 13272007
Where
Industrial site
When
September 2026
Test
The field at the site, checked against a reference magnetometer alongside
Data
The raw records are kept, and are not shared
AGS-M1 · REV B

Gates

7 OF 9 CLOSED
  • Field-testedSeptember 2026 DONE
  • Bench board accepted2026-07-09 DONE
  • Noise sweep, sensor attached2026-07-10 DONE
  • In-spec clock tree2026-07-11 DONE
  • Runs from an external clock, clock-following firmware2026-07-11 DONE
  • Output rates 10 / 50 / 100 / 200 S/s DONE
  • Photographs of a built board2026-11-01 DONE
  • Datasheet AGS-DS-001 Rev A ACTIVE
  • Noise re-run with a shielded short OPEN
LOG

Milestones, newest first.

  1. REV C Next revision in design. The clock tree and pin map for the next board revision are drawn up. Nothing is built yet. IN DESIGN
  2. REV B Runs from an external 8 MHz clock. On the clock-following firmware, which we send on request, the bench board ran entirely from an external 8 MHz clock and passed all four checks: it answered with no clock present, locked and sampled once a clock was applied, followed the clock through a 1 % shift, and stopped cleanly when it was lost. No hardware change was needed. The standard firmware runs the board from its own oscillator. MEASURED
  3. REV B In-spec clock tree. A firmware change brought a peripheral clock inside the processor's rated limit, with no change to the board. The serial link, a 200 S/s stream and an on-board recording cycle were checked again afterward and ran as before. MEASURED
  4. REV B Chain noise separated from the environment. With the sensor disconnected and its inputs shorted at the sensor connector, the noise was 15 to 73 times lower than in the sensor-attached sweep, and it fell steadily as the converter rate came down. At 503.8 S/s it read 0.035 nT/√Hz with the standard deviation divided by √(converter rate ÷ 2), or about 0.11 divided by √(output rate ÷ 2). MEASURED
  5. REV B Noise sweep, sensor attached. With the fluxgate attached, a sweep across converter rates suggested that the bench's magnetic environment, not the board, set the noise. The shorted-input run the same day confirmed it. MEASURED
  6. REV B Bench board accepted. Every command of the board's original firmware passed over the RS-485 link. MEASURED
NOT YET CLAIMED

What this site does not say yet, on purpose.

  • A board mass or an operating temperature range

    Neither is measured or specified yet, so neither is given.

  • Several boards on one clock

    One board has run from an external clock, on the clock-following firmware. A clock shared across boards has not been tested.

  • A noise plot

    The shielded-short re-run is open. Its plot, and the method it settles, go on the site when it reports.

  • Accuracy in your installation

    What a board reads depends on what sits beside it. No application accuracy figure is published.